Lamar Owen wrote:
[EMAIL PROTECTED]">responsiveness != performance IT works OK for a low number of concurrent users/processes to increase percieved performance, but to get real gains on large systems with large numebrs of users and processes you actually decrease the responsiveness of individual tasks (IE make the system a little less likely to context switch or pre-empt) and schedual in batches or clusters rather than one-at-a-time. For a desktop/workstation this would be insane, and drive a user to kill someone, but for systems that handle several hundred users (interactive or not) this improves overall perfomance.The low-latency and preemptible patches are not meant for performance
gains, but for responsiveness, and are not designed to be used in servers,
only in workstations/desktops.
ISTM that improving interactive performance would also improve multiuser
performance in a server, as low latency and kernel preemption can increase
multiuser server responsiveness.
2.4.18 has a lot of work done to the VM, but most importantly has work done to the queue elevator code, thats probably whats doing most of the work (throttling big writers) of seeing better overall system performance.
